C a r d i o M i n d
Read faster. Decide as you always have.
A workspace for 12-lead ECGs. Open a case, read what the model found and why, measure the intervals yourself, record your interpretation, export the report. The analysis does the legwork — the read stays yours.
The Platform can measure a trace in seconds. Only you can take responsibility for the read.
Everything arrives editable
The interval measurements, the detected condition, the drafted report — each is a proposal to confirm, adjust, or overrule. A finding becomes your interpretation at the moment you review it and save it, and not before.
Every claim shows its working
Each finding carries the model's confidence and a priority. Each measurement is coloured against published thresholds. Where a number can't be trusted, the analysis says so and explains why rather than quietly reporting it.
The one rule. Never sign off a finding you haven't verified against the trace yourself. A clear result means the model didn't recognise one of the fifteen conditions — it is not a guarantee that the ECG is normal.
Send an ECG and six things happen before you see it. You manage none of them. It is still worth knowing what the results rest on.
An advanced custom deep-learning network, trained on labelled 12-lead ECGs, reading the signal four ways at once.
Not a fixed set of if-then rules. It reads a full 12-lead digital signal where one exists; from a printed or scanned page it gets each lead for the 2.5 seconds that page prints it in, plus the rhythm strip throughout. Either way, four angles at once.
Detection is learned,
not threshold-based
The model decides what a trace shows from the waveform itself. The numeric thresholds you see further down this page — PR ≥ 200 ms, QRS ≥ 120 ms — are not how conditions are detected. They are applied after detection, to colour the measurements. Detection and colouring are separate steps. There is also a dedicated component for the clinically awkward atrial fibrillation versus atrial flutter distinction.
The workspace you sign in to, running here on synthesised signal.
Pick any of the fifteen conditions. The twelve leads, the metric cards, the annotation tree and the drafted reasoning all follow it — and each column carries a different 2.5-second window, exactly as a printed page hands them over. The measurement tools are drawn but inert: select one and it will tell you what it does on a real case.
What you get on sign-in-
Split the editor
Drag a tab to the left or right edge and hold a new recording against a prior one on the same screen.
Ctrl+Bhides the panels for more trace. -
Calipers and cursor
Measure any interval by hand — successive calipers cycle through six colours so several stay readable — or track lead, time and amplitude live under the pointer.
-
Read it your way
Zoom and pan, gain at 5, 10 or 20 mm/mV, vertical spacing counted in ECG big cells, the 3 × 4 grid or a single stacked column, and exactly the leads you choose.
-
Ask the assistant
Questions about the open case are answered from its own analysed data, so a known question always returns the same validated figure rather than a fluent guess.
-
Work a worklist
Quick-filter chips, or a precise query —
has:afib rate>110 priority:critical— with multi-column sorting and the analysis queue tracked as it runs. -
Sign and export
Edit the drafted report, rate its accuracy, then generate a clinical PDF carrying the trace, your measurements, your settings and your notes. All of it audited.
This workspace is live. Pick any of the fifteen conditions above and the twelve leads, the metric cards and the annotation tree all follow it; switch the overlays on and off in the annotations panel; open a measurement's ⓘ for the special notes behind it; and click any R peak on the Lead II reference strip to open the beat plot, which superimposes every beat in the recording against the averaged beat shape. The measurement tools are shown but not live here — select one and it will tell you what it does in the app.
Highly optimised for mobile. Review an ECG anywhere, at any time.
The workspace is the same on a phone as it is on a reporting station. Not a cut-down mobile view with the hard parts removed — the same twelve leads, the same measurements, the same annotation layers, laid out for the screen you happen to be holding.
The screen goes to
whatever you are reading
The pattern behind all three is the same: panels yield to the trace. In portrait the metric row keeps its cards at full size and scrolls, because a card squeezed to half width is a card you cannot read. Turn to landscape and the metric panel unpins itself rather than competing for height. Fullscreen takes it further and clears everything. At no point does the trace get smaller so that a panel can stay put — and the measurements are never more than one tap away.
Open. Read. Measure. Record.
Every review has the same shape, whatever the case turns out to be. Four steps, in one workspace, in a browser tab you can split against a prior recording.
CardioMind is built by Intelligent Mind Labs. Clinical AI is not a demo: it earns its place by being measured, explained, auditable, and answerable to the clinician holding the pen.
The same conviction runs through the whole product. Detection is learned, so it can read a waveform rather than a rulebook — but the reasoning is grounded in a curated ECG knowledge base rather than left to invent itself, and the assistant answers from the case's own analysed data so a known question always gets the same dependable answer. Where a measurement can't be trusted, the framework attaches a caveat and the explanation reflects it. Nothing is asserted more confidently than the signal allows.